A novel non-destructive approach for cashmere authenticity: Combining spectroscopic techniques and machine learning...

Area of Science:

• Textile science
• Analytical chemistry
• Chemometrics

Background:

• Cashmere is a luxury fiber frequently adulterated with sheep-down.
• Current identification methods are subjective, costly, or time-consuming.
• Spectroscopy combined with machine learning offers a potential solution for accurate authentication.

Purpose of the Study:

• To develop an accurate and efficient method for distinguishing pure cashmere from sheep-down and their blends.
• To evaluate the performance of near-infrared (NIR) and ATR-FTIR spectroscopy coupled with various machine learning algorithms.
• To establish a reliable technique for cashmere authentication and combat adulteration.

Main Methods:

• Analysis of 200 cashmere and 200 sheep-down fiber samples using NIR and ATR-FTIR spectroscopy.
• Spectral data processing with principal component analysis (PCA), partial least squares-discriminant analysis (PLS-DA), data-driven soft independent modeling of class analogy (DD-SIMCA), and convolutional neural networks (CNNs).
• Simulation of blended samples (10-50% sheep-down) using a linear mixing model.

Main Results:

• PLS-DA and DD-SIMCA models showed limited accuracy (41-73%), with ATR-IR models generally outperforming NIR.
• The CNN model, particularly with ATR-IR data, achieved perfect classification (100% accuracy, sensitivity, and specificity) for pure and blended samples.
• The ATR-IR-CNN approach significantly outperformed traditional chemometric methods.

Conclusions:

• The ATR-IR-CNN approach provides a rapid, non-destructive, and highly accurate method for cashmere authentication.
• This technique offers a robust solution to combat adulteration, protect consumers, and improve textile industry quality control.
• The developed method demonstrates superior performance compared to existing techniques for fiber authentication.